Technology vendor of the year: SS&C Algorithmics

Technology has, arguably, never been sexier or more glamorous. Artificial intelligence stocks are flying high, and the questions AI poses are being debated in homes, bars, boardrooms and parliaments around the world.

So, it’s refreshing to hear Curt Burmeister emphasising nuts-and-bolts practicality. To be clear, the chief technology officer (and co-head) of SS&C Algorithmics has some AI wizardry to show off as well, but a lot of his focus is about getting more out of what’s already there.

This approach seems to be paying off. In this year’s Markets Technology Awards, the firm scooped three closely related prizes – for market risk management, counterparty risk and XVAs. Judges described SS&C Algorithmics as “technologically strong”, “technically solid” and “performant”. 

Or, to borrow a famous tagline from a British manufacturer of varnishes and wood stains: it does exactly what it says on the tin.

And in recognition of this success, SS&C wins this year’s technology vendor of the year.

Everyone embraced the cloud three to five years ago, but the next step is to start managing your cloud software with a cloud mindset

Curt Burmeister, SS&C Algorithmics

As an example of the company’s focus, Burmeister points to cloud technology – not a new trend, he acknowledges, but one that isn’t yet being used to its full potential. SS&C Algorithmics recognised this during development projects.

“We’d get a client request for a new feature, so the developers would build it, and release a new version, and then somebody would install it and configure it for whatever client setup they wanted to test, and then they’d test it. Altogether, it might take a few days for us to deploy. And we said, ‘Hang on. This makes no sense. We should be able to deploy our own application in a few minutes’,” says Burmeister.

The company switched to a new approach, using GitHub and GitHub Actions, to manage client-specific configurations and deploy software. GitHub is commonly used to manage the development of new features, and applying the same approach for client configurations offered a way to automate the entire deployment process. Developers are now able to automatically deploy a development build and configure it in a matter of minutes, a task that previously took days, Burmeister says. “Everyone embraced the cloud three to five years ago, but the next step is to start managing your cloud software with a cloud mindset. And this is just one example,” he says.

Dormant chips

A similar spirit animates the firm’s HiPER risk engine – the kit that enables users to run speedy instrument and portfolio simulations. Some vendors have tried to accelerate these calculations by using graphics processing units – special chips optimised to run large numbers of specific operations.

SS&C Algorithmics opted for vectorisation instead, a capability that is lying dormant in most standard chips. Rather than running a series of calculations sequentially, it allows a block of calculations to be run simultaneously.

“Typically, it’s eight double-precision floating-point numbers at a time – you can add or multiply eight doubles at once – as opposed to adding the first two elements, then the second two, and so on. Those vector instructions are on every Intel chip, every AMD chip out there. But very few applications use those instructions,” says Burmeister.

The firm considered using GPUs, but the chips have become increasingly costly. Vectorisation could be deployed with no additional cost – just the time required to rewrite the software so it could take advantage of en masse calculations.

If you can reduce an eight-hour batch process to a 20-minute sprint, then you can do it four or five times a day. People aren’t there yet. But they’re starting to see the potential

Curt Burmeister, SS&C Algorithmics

The firm is also using adjoint algorithmic differentiation (Burmeister prefers to call it simply ‘adjoint differentiation’). This maths ‘trick’ allows calculations to skip the traditional but time-consuming bump-and-revalue process when calculating derivatives sensitivities – again, more of the work is done simultaneously rather than sequentially.

The end result is a simulation that can be as much as 1,000 times faster than the firm’s legacy simulation engine, says Burmeister – but is more commonly two-to-four hundred times faster.

This is important, because the kind of heavy-duty calculations that HiPER is designed for are an increasingly important part of the markets business – embedded in margin calculations, derivatives valuation adjustments and incoming trading book rules.

And this is the point, Burmeister says – speed is not an end in itself. It’s a way of allowing market participants to do more: price faster, tackle more complex problems, assess risk more often.

“Normally, in computer science, if you can make a 25% improvement, that’s enormous. But we’re talking hundreds of times faster – up to 1,000 times. It’s really, really powerful. And it means you can run the same calcs really fast for the same cost – or you can take advantage of those savings to do more. If you can reduce an eight-hour batch process to a 20-minute sprint, then you can do it four or five times a day. People aren’t there yet. But they’re starting to see the potential,” he says.

One regional bank client of SS&C Algorithmics demonstrates what can be done. The bank has moved from calculating 10 XVA sensitivities to calculating 10,000 – and, Burmeister says – using a fraction of the hardware they previously required.

“It’s really, really impressive. And you might ask, ‘Well, why aren’t people jumping all over this?’ I think they will over time,” he says.

Global reg, Chinese policy

Another thing that might change over time is adoption of internal models for calculating trading book capital. The new capital regime – the Fundamental Review of the Trading Book – has been implemented this year in several jurisdictions, including Canada, Japan and Switzerland, with a grand total of one bank so far opting for the more sensitive, but more burdensome modelling approach.

Implementation in the EU and UK is expected to hinge on when and how the US adopts FRTB. But there are signs that some of these banks are keeping their options open – potentially starting life under the cruder, standardised approach, then migrating to the internal models approach later on. Burmeister says at least two clients are using SS&C Algorithmics’ software to pursue compliance with the IMA as well as the standardised approach.

“More people chose the standardised approach than I expected, across the board. Now, we have picked up signals that people are re-evaluating this, and that maybe people who initially opted for standardised are going to change to the IMA,” he says.

In China, the challenge for SS&C Algorithmics – and other foreign analytics vendors – has been policy, rather than regulation. The country’s Xinchuang initiative is an attempt to ensure critically important industries are not dependent on third-country technology – potentially a way of insulating China from pressure in the event of some future conflict.

Banking is one of the industries that is subject to Xinchuang, which creates a challenge for overseas vendors of trading and risk systems – it means those systems have to be compatible with China’s emerging domestic technology stack. But these expectations have not been laid out in a detailed regulatory or technical document, leaving vendors to work through the implications themselves.

“This hit us first two years ago, when one of our largest clients said, ‘Hey, we’d like you to run the application on this Chinese version of Linux’. It was a very challenging business problem,” says Burmeister.

SS&C Algorithmics responded by striking a deal with a local partner, Tansun Technology, in December last year. Under the deal, the clients were transferred to Tansun – the local partner now provides support to those clients – and SS&C Algorithmics has now built a new version of its market risk software that is compatible with domestic databases and operating systems, says Burmeister.

“We delivered that in September. It’s a way of allowing our clients to have continuity. It was pretty clear to us we were going to lose our 13 clients over time if we did not find a local partner,” he says.

Other vendors have done something similar. Risk.net has previously reported that Numerix and Finastra have both struck partnership deals with domestic Chinese technology firms, to help them support their products locally. Other foreign vendors are said to have walked away from the market altogether.

SS&C Algorithmics is aiming to be there for the long run.